54 



THE INTERFEROMETRY OF 



presence of fringes, and of the opaque mirror N as specified. Of these, e M is 

 systematically larger than e N for reasons which do not appear. The screws 

 were of American and foreign make, but they could not be so different. It 

 is due very probably to residual curvature in the mirrors and surfaces, whereby 

 fringes on the left (N) vanish sooner than those on the right (M) . The datum 

 y is the displacement of the right-angled reflecting prism P f , parallel to the 

 component rays bb'. This value is smaller than e for reasons already discussed 

 1 6. All measurements were frequently repeated and the means finally 



in 



taken for comparison with dd/d\. 



In the third series (ruled grating and concave grating) with specially brilliant 

 spectra, the phenomenon of figure 32 was observed. A wide field of faint 

 fringes was visible, enormously accentuated and clear in the narrow strip of 

 the linear phenomenon. As the micrometer mirror at M moves forward, these 

 faint fringes shift bodily across the stationary bright linear strip, beginning 

 therefore with the pattern a and ending with b. The faint fringes follow the 

 rules of displacement interferometry. 



a h 



32 



OH 



33 



4&d0/d& 



4-\ 34 



4- 8 1% 16 



In addition to the data of the table, a large number of miscellaneous tests 

 were made with the reflecting prism in different positions. Unless brought 

 too far to the rear, when the beams are lost at the edge and e too small, the 

 results for fine and coarser fringes were of the same order. 



The values of y have been graphically given in figure 33 ; those for e are 

 not sufficiently regular in the dispersive powers above i ,000 for this treament. 

 It is probable, for instance, that at 16,900 the sliding along the prism surface 

 is interfered with. All the data, in consideration of their limitations, bear out 

 the inference that the range of displacement within which fringes are seen 

 increases in marked degree with the dispersion. The average initial ratio 

 2\/(d6/d\) is about 60X10"* cm. 



A very surprising result in these experiments is the efficiency of the film 

 grating in series IV and V, not only in the first but in the second order of 

 spectra. 



